[hibernate-dev] Snapshot conflict with merge

[Max Rydahl Andersen]

Hi,

The challenge is that both have merit.

The first one (dirty merge + dirty attached = exception) is more correct  
 from a consistency perspective.
The second one (dirty merge + dirty attached = dirty merge wins) wins is  
(probably) usefull in many practical cases (and works like I think most  
people thing merge works ...that you
are telling hibernate that the object to be merged in is the most correct  
one).

As far as I remember from the last time we discussed it we considered  
adding a MergeMode to
allow users to enable the "dirty merge wins" strategy.

The spec should list which one wins otherwise in this merge situations  
otherwise merge's behavior is non-specfied across implementations which  
will lead to alot of problems.

Have you tested it both in hibernate core and against entitymanager ? Do  
they behave differently ?

/max

> I wrote (tested and verified with Gavin):
>
> Finally, merging doesn’t allow the following scenario, where two  
> modifications conflict on one snapshot:
>
> item.getId() // The database identity is "1234"
> item.setDescription(...);
>
> Session session= sessionFactory.openSession();
> Transaction tx = session.beginTransaction();
>
> Item item2 = (Item) session.get(Item.class, new Long(1234));
> item2.setDescription(...);  // Not allowed!
>
> return session.merge(item); // Throws exception!
>
> tx.commit();
> session.close();
>
> The detached object’s description is modified as well as the persistent  
> object’s. There is no way for Hibernate to determine who will win in  
> this situation and a StaleObjectStateException is thrown when you try to  
> merge() onto a dirty persistent instance.
>
> Emmanuel comments:
>
> No exception is raised in my tests. The merged state takes precedence  
> over the initially loaded and modified state (regardless the object has  
> been versioned or not) AFAIR, the spec does not talk about such an  
> exception either.
>
> What is it?
>
>
> _______________________________________________
> hibernate-dev mailing list
> hibernate-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hibernate-dev



-- 
--
Max Rydahl Andersen
callto://max.rydahl.andersen

Hibernate
max at hibernate.org
http://hibernate.org

JBoss a division of Red Hat
max.andersen at jboss.com